Re: [Idr] I-D Action: draft-ietf-idr-error-handling-03.txt

[Jakob Heitz <[e-mail]>]

On Dec 9, 2012, at 4:12 PM, "Chris Hall" <[e-mail]> wrote:

> Jakob Heitz wrote (on Sun 09-Dec-2012 at 23:37 +0000)
>> IMO, another goal is not to require any change to the peer.
>> Not even a little bit.
> 
> Sure.  It would be good to be able to improve error handling
> unilaterally.  As discussed elsewhere, I think it is possible to do
> that, subject to some limitations.
> 
> Without those limitations, the receiver is at risk of applying
> "treat-as-withdraw", but failing to identify all NLRI in the message,
> and hence continuing with some invalid and/or out of date routes.  IMO
> that is best avoided.  There may be a good argument for rejecting the
> limitations and accepting the risk of some invalid and/or out of date
> routes -- I look forward to considering it.  
> 
>> Changing the peer behaviour (even a little bit)
>> is an entirely different story.
> 
> Hmmm.  Section 3 of the draft states:
> 
>  "To facilitate the determination of the NLRI field
>   in an UPDATE with a malformed attribute, the
>   MP_REACH_NLRI or MP_UNREACH_NLRI attribute (if
>   present) SHALL be encoded as the very first..."
> 
> which looks like a change in peer behaviour to me... but my eyesight
> is not what it was ?

That is not required. It is a SHALL. If the peer does it, great. If not, nothing broke.
It's different to requiring a protocol change.